一种快速的Rough集属性约简遗传算法

Efficient Genetic Algorithm of Attribute Reduction Based on Rough Set

遗传算法适合复杂问题的处理因此可用于属性约简的求解.目前利用遗传算法进行属性约简的主要不足是:适应度函数计算复杂,效率不高.尤其在处理大型决策表时,计算时间将大量聚集在适应度函数的计算上,从而导致算法性能下降.为了更快的计算适应度函数,在研究基于正区域的区分对象对集的基础上,设计了一种计算适应度函数的快速方法.利用启发信息设计了一种快速的属性约简遗传算法.通过实例分析和算法实验表明该算法能够高效求出决策表的属性约简并且适合处理大型决策表.

Genetic algorithm is adapted to deal with complicated problem, so it often computed the attribute reduction in rough set. But at present, by making use of genetic algorithm to compute attribute reduction, the major drawback is the multiple computation for the fitness function, so as to the efficiency is not high. In particular when dealing with a large decision table, the multiple computa- tions gather in the computation value of the fitness function, so the algorithm performance is not good. In order to get the value of the fitness function quickly, in this paper, on the condition of discernibility object pair set based on positive region, a new quick method of computing the fitness function is designed, at the same time, a new efficient genetic algorithm to attribution reduction is presented by using heuristic information. Finally, an emulate example and experiment results illustrate the efficiency of the new genetic algorithm to compute the attribute reduction of the decision table especially tackling a large decision table.